PrivateEye: Scalable and Privacy-Preserving Compromise Detection in the Cloud


Behnaz Arzani, Microsoft Research; Selim Ciraci, Microsoft; Stefan Saroiu, Alec Wolman, and Jack Stokes, Microsoft Research; Geoff Outhred and Lechao Diwu, Microsoft


Today, it is difficult for operators to detect compromised VMs in their data centers (DCs). Despite their benefits, the compromise detection systems operators offer are mostly unused. Operators are faced with a dilemma: allow VMs to remain unprotected, or mandate all customers use the compromise detection systems they provide. Neither is appealing: unprotected VMs can be used to attack other VMs. Many customers would view a mandate to use these detection systems as unacceptable due to privacy and performance concerns. Data from a production cloud show their compromise detection systems protect less than 5% of VMs.

PrivateEye is a scalable and privacy-preserving solution. It uses sanitized summaries of network traffic patterns obtained from the vSwitch, rather than installing binaries in customer VMs, introspection at the hypervisor, or packet captures. The challenge it addresses is protecting all VMs at DC-scale while preserving customer privacy using low-signal data. We developed PrivateEye to meet the needs of production DCs, and our data collection agent is deployed across all DCs of a large cloud. Evaluation on VMs of both internal and customer VM's shows it has an area under the ROC curve -- the curve showing the model's true positive rate vs its false-positive rate -- of 0.96.

NSDI '20 Open Access Sponsored by NetApp

Open Access Media

USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.

@inproceedings {246284,
author = {Behnaz Arzani and Selim Ciraci and Stefan Saroiu and Alec Wolman and Jack Stokes and Geoff Outhred and Lechao Diwu},
title = {{PrivateEye}: Scalable and {Privacy-Preserving} Compromise Detection in the Cloud },
booktitle = {17th USENIX Symposium on Networked Systems Design and Implementation (NSDI 20)},
year = {2020},
isbn = {978-1-939133-13-7},
address = {Santa Clara, CA},
pages = {797--815},
url = {},
publisher = {USENIX Association},
month = feb

Presentation Video